Scholarlink

<p>Amyotrophic Lateral Sclerosis (ALS) is a neurodegenerative disease that involves the degeneration of motor neurons, resulting in muscle weakness and progressive paralysis. Early diagnosis of ALS remains difficult due to its multifaceted pathophysiology and varied symptoms. This research introduces NeuraSpeech, a new hybrid model that integrates state-of-the-art machine learning algorithms and blockchain technology for ALS detection using speech analysis and secure data exchange. We introduce a multi-modal feature extraction method that combines conventional acoustic features, deep learning representations, non-linear dynamics measures, and multi-resolution analysis. A two-stage feature fusion and selection process selects an optimal subset of 22 features with high discriminative power. Our ensemble classification system integrates conventional machine learning models and deep neural networks using weighted voting. The method attains 97.2% accuracy with ±1.0% standard deviation, showing robust performance across patients. The integration of Hyperledger Fabric on blockchain offers data integrity, privacy, and secure sharing of diagnostic reports among healthcare institutions with 28-55 TPS throughput and 0.9−1.9&#xa0;s latency. In this paper, the methodological process and results are presented, showing the ability of NeuraSpeech to improve the early diagnosis of ALS while maintaining the security and privacy of patient information.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Neuraspeech: a secure machine learning-blockchain framework for speech-based amyotrophic lateral sclerosis detection

  • 期刊论文

  • 发表日期：

    2026年5月6日

查看全文
  • Ayoub Louja,
  • Yassin Zaiouane,
  • Manal Benchrif,
  • Najoua Azizi,
  • Abdellah Jamali,
  • Najib Naja

摘要

Amyotrophic Lateral Sclerosis (ALS) is a neurodegenerative disease that involves the degeneration of motor neurons, resulting in muscle weakness and progressive paralysis. Early diagnosis of ALS remains difficult due to its multifaceted pathophysiology and varied symptoms. This research introduces NeuraSpeech, a new hybrid model that integrates state-of-the-art machine learning algorithms and blockchain technology for ALS detection using speech analysis and secure data exchange. We introduce a multi-modal feature extraction method that combines conventional acoustic features, deep learning representations, non-linear dynamics measures, and multi-resolution analysis. A two-stage feature fusion and selection process selects an optimal subset of 22 features with high discriminative power. Our ensemble classification system integrates conventional machine learning models and deep neural networks using weighted voting. The method attains 97.2% accuracy with ±1.0% standard deviation, showing robust performance across patients. The integration of Hyperledger Fabric on blockchain offers data integrity, privacy, and secure sharing of diagnostic reports among healthcare institutions with 28-55 TPS throughput and 0.9−1.9 s latency. In this paper, the methodological process and results are presented, showing the ability of NeuraSpeech to improve the early diagnosis of ALS while maintaining the security and privacy of patient information.